[PATCH v3] drm: don't run atomic_async_check for disabled planes

[Xaver Hugl]

> As I commented earlier in the gitlab issue [1], any change of property, including disabling a plane is not allowed in the async commit.
> We must disable a plane (e.g. HW cursor) during the first (synchronized) flip, and allowing later flips to proceed asynchronously.
> This change should be done in the compositor.
No change is needed there, compositors already do that.
> As per Ville's opinion in related series [2], kernel driver should reject all these disabled
> Planes in the drm core and driver should only get the planes which is supported with async flip. Based on his comment, I have started
> Working and will be addressing it in the next version of my series [3].
As long as it only filters out planes that were and still are
disabled, I think that could work out fine - in theory they shouldn't
have any side effects.
Note though that my patch intends to specifically fix a regression for
amdgpu in 6.15, which I think we can do with fewer changes.

> [1]. https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/13834#note_2994595
> [2]. https://lore.kernel.org/all/aHAg2J-uFLLWINqp@intel.com/
> [3]. https://patchwork.freedesktop.org/series/151280/
>
> Regards,
> Naveen Kumar